Power amplifiers are used within many communication apparatuses in order to raise to a high level the power of a signal that is to be transmitted. These power amplifiers can be designed to operate within a number of different classes, the class of operation referring to the conduction angle of the transistor(s) within the amplifier. The difference in conduction angle results in different output waveforms when an input signal is applied. The most important result of these different classes of operation concerns their affect on the linearity and efficiency of the power amplifier. In fact, one should understand that there is a trade-off between linearity and efficiency within amplifiers such that when linearity increases, efficiency decreases and vice versa.
Amplifiers that operate in Class A, those with a conduction angle of 360.degree., typically have good linearity, that being constant gain and delay over a wide range of frequencies and power, and low distortion, that being low harmonics and intermodulation products generated by the amplifier. The disadvantage of these Class A amplifiers is their poor efficiency. Theoretically, an amplifier operating in Class A can at most have a 50% efficiency. A Class A amplifier can be operated from almost zero output power to some maximum output power known as the saturation power. At very low levels of output power, the distortion is typically not measurable. As the output power approaches saturation power, the distortion increases and becomes very large when the saturation power is reached. Because a Class A amplifier uses almost a constant amount of DC supply power, the efficiency is almost zero at very low output powers and typically only approaches 50% when the distortion becomes significant.
Amplifiers operating in Class B, those having a conduction angle of 180.degree., have increased efficiency but a lower level of linearity than Class A. The maximum theoretical efficiency possible with this class of amplifier is 78%. The difficulty is that a Class B amplifier has distorted output, especially when the input signal is close to zero volts. Further classes of amplifiers (Classes C through F) with smaller conduction angles (&lt;180.degree.) are possible. These amplifiers each have further increased efficiencies at the expense of more distortion.
A compromise amplifier, that has lower distortion than a Class B amplifier but approximately the same efficiency, is a Class AB amplifier. Amplifiers operating in Class AB have a conduction angle larger than 180.degree., but lower than 360.degree.. These amplifiers avoid some, but not all, of the distortion problems of the Class B amplifiers while maintaining most of the Class B amplifier's efficiency.
Techniques currently exist to reduce the distortion within Class AB amplifiers so that even applications that require high linearity can utilize them. These characteristics make Class AB amplifiers particularly appropriate for applications in which a reasonably high efficiency and low distortion are required. These same techniques can be used to reduce the distortion in Class A amplifiers operating near saturation power.
Some applications in which Class AB amplifiers have been previously implemented include communication apparatuses such as base stations. Some of the key factors that contribute to the overall cost and performance for base stations are the size, cost, and quality of their power amplifiers. Without the use of techniques to lower distortion as mentioned herein above, the quality of the output signals from Class AB power amplifiers and Class A amplifiers driven near saturation, is not satisfactory for base station applications. The use of Class A amplifiers at power levels significantly below saturation power, on the other hand, is typically unacceptable due to low efficiency.
One technique to reduce the distortion within an output signal of an amplifier 20 is the use of a feed forward error compensation path 22 as depicted within FIG. 1. In simplistic terms, such a path 22 generates an error signal by comparing a signal being input to an amplifier 20 and that being output from the amplifier, and then amplifying and filtering the result of the comparison to obtain the correction signal. By then adding the correction signal to the output of the amplifier 20, the error within the output signal (i.e. the distortion) as a result of the amplifier 20 can be reduced. The distortion is reduced because the distortion components in the correction signal are in antiphase with the distortion components in the signal at the output of amplifier 20 and of equivalent power to them after the addition process.
The delay caused by this feed forward path 22 is one of the key problems with this technique, as the correction signal can have the desired phase relationship with the output signal from the amplifier 20 only over a narrow frequency range; outside this range there is a phase error. This "phase error" decreases the effective correction that is obtained with use of the feed forward error compensation path. As the phase error increases, the error correction decreases to an extent such that when the phase error is at 60.degree. there is no correction. If the phase error exceeds 60.degree., the "correction" makes the distortion worse.
The current method to overcome this delay problem is to use a delay line between the output of the amplifier 20 and the location in which the amplified error signal is being added to the output signal. If the length of the delay line is selected properly, the delay caused by the delay line will be equal to the delay caused by the feed forward error compensation path 22. There are a number of key problems with a setup such as this with a delay line. Firstly, the delay line is costly and relatively large, thus significantly increasing the cost and size of the overall power amplifier. Further, the delay line is implemented within a high power area of the power amplifier and results in attenuation of the amplified signal. This attenuation not only reduces the output power of the overall power amplifier requiring a larger amplifier 20 to get the same power, but also causes considerable heat dissipation. This heat dissipation must be compensated for with large and expensive heat sinks or otherwise, the resulting signal will have changes in phase and gain that lead to an even greater phase error.
An alternative to the delay line described above is a delay filter being used between the output of the amplifier 20 and the location in which the subtraction of the error signal occurs. Although this delay filter doesn't reduce the overall power or increase the overall size of the power amplifier as much as the delay line, these areas are still of concern. The amount of heat dissipated within the delay filter will be significant resulting in similar problems to that described for the delay line. One key disadvantage of using the delay filter is the components that must be used for this purpose. Since the delay filter is implemented at a high power area of the power amplifier, the delay filter would have to be implemented with large waveguide or discrete components, making the filter expensive, large and difficult to manufacture. These delay filters are simply impractical in cost-conscience power amplifier designs.
A second technique to reduce the distortion within an output of an amplifier 20 is to use a feedback error compensation path 24 as depicted within FIG. 2. With this technique, the input and output to the amplifier 20 are compared and the results are used to adjust the input to the amplifier 20. Many variations of this technique are possible. For example, the results can be proportional to magnitude and phase error and hence connected to a magnitude and phase adjuster. Alternatively, the results can be proportional to in phase and quadrature amplitude error, and hence connected to an in phase and quadrature amplitude adjuster. FIG. 2 shows two feedback paths, though in general there can be one or a plurality of feedback paths.
One key problem with the use of a feedback path 24 is the possibility of oscillations within the output signal due to the delay within the path itself. The delay results in a lag between the signals being compared and the error signal adjusting the input signal to compensate for distortion. The ideal phase shift within the error signal is 180.degree. so that it subtracts the error signal from the input signal S.sub.IN (t). As the frequency increases, the lag increases the phase shift which eventually approaches a threshold near 360.degree.. At this frequency, the output signal will oscillate if the loop gain is greater than one. These oscillations represent unacceptable instability. Filters are required to reduce the loop gain to less than unity before the phase threshold is reached. These filters also reduce the bandwidth over which correction is effective. Hence, this feedback technique has a significantly limited bandwidth for it to be used effectively, essentially limiting the bandwidth for the entire power amplifier to operate properly.
As described above, each of the currently well-known techniques to reduce the distortion within Class AB amplifiers implemented within base stations have cost, size, and/or bandwidth problems. Therefore, these techniques for reducing distortion must be improved to limit these problems so that a cheaper, smaller, higher efficiency, and/or higher bandwidth power amplifier can be developed.